3 ms·
> How do we know the AI output is accurate? From the blog post where they announced 2023 grand prize: they did a couple of things to verify the results, includ
by omoikane 2y ago
> How do we know the AI output is accurate?
From the blog post where they announced 2023 grand prize: they did a couple of things to verify the results, including scanning the same area multiple times and making sure that multiple models produce similar results.
https://scrollprize.org/grandprize#how-accurate-are-these-pictures https://scrollprize.org/grandprize#how-accurate-are-these-pi...
- luma 2y agoThey also created their own scroll, cooked it, and scanned it to create a ground truth dataset.
- mmooss 2y agoI'm sure they work to improve the accuracy of the output but it's not 'truth' unless it's the same data - which of course is impractical. That is, unless 'truth' means something different in ML, which would be troubling!
- luma 2y agoGround truthing does in fact mean something different in ML but I’m not clear on how that is troubling.
- mmooss 2y agoIf I understand you, AI/ML is redefining "truth" from facts of reality to whatever the input (or maybe the internal state) says. It's corrupt, a lie, and will obviously output inaccuracies - but only inaccurate by reality's standard!